[
https://issues.apache.org/jira/browse/HBASE-21005?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16573757#comment-16573757
]
Sean Busbey commented on HBASE-21005:
-------------------------------------
{quote}
bq. why aren't we just publishing this jar as a part of hbase-thirdparty?
Because I could turn this around in ~2-3 hours and it is ultimately still
"throwaway". I didn't have motivation to spend time putting it into
hbase-thirdparty.
{quote}
fair cop.
{quote}
bq. also please link to https://github.com/apache/maven-fluido-skin/pull/1
and/or MSKINS-137 so we have some hope of getting back to basic artifacts at
some point.
Where do you want links to these?
{quote}
I would have said {{src/site/site.xml}}, but maybe just having it as a comment
here is enough since we already reference this jira at the place where I'd call
out "remove this after MSKINS-137 is fixed"? I have talked myself down to "nit"
status. :)
> Maven site configuration causes downstream projects to get a directory named
> ${project.basedir}
> -----------------------------------------------------------------------------------------------
>
> Key: HBASE-21005
> URL: https://issues.apache.org/jira/browse/HBASE-21005
> Project: HBase
> Issue Type: Bug
> Components: build
> Affects Versions: 2.0.0
> Reporter: Matt Burgess
> Assignee: Josh Elser
> Priority: Minor
> Attachments: HBASE-21005.001.patch
>
>
> Matt told me about this interesting issue they see down in Apache Nifi's build
> NiFi depends on HBase for some code that they provide to their users. As a
> part of the build process of NiFi, they are seeing a directory named
> {{$\{project.basedir}}} get created the first time they build with an empty
> Maven repo. Matt reports that after a javax.el artifact is cached, Maven will
> stop creating the directory; however, if you wipe that artifact from the
> Maven repo, the next build will end up re-creating it.
> I believe I've seen this with Phoenix, too, but never investigated why it was
> actually happening.
> My hunch is that it's related to the local maven repo that we create to
> "patch" in our custom maven-fluido-skin jar (HBASE-14785). I'm not sure if we
> can "work" around this by pushing the custom local repo into a profile and
> only activating that for the mvn-site. Another solution would be to publish
> the maven-fluido-jar to central with custom coordinates.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
